Radio frequency power supply control method and device, electronic equipment and storage medium

Abstract

The invention belongs to the technical field of radio frequency power supply control, and discloses a radio frequency power supply control method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining the operation parameters of a radio frequency power supply, obtaining the historical operation parameters of a power supply database and the corresponding historical state data, and carrying out the control of the operation parameters through Kalman filtering and a model prediction control algorithm, according to the historical operation parameters and the corresponding historical state data, constructing a model prediction control algorithm model, inputting the operation parameters into the model prediction control algorithm model, calculating to obtain prediction state data of the radio frequency power supply, and based on a preset objective function, combining the prediction state data with the actual state data of the radio frequency power supply to obtain a prediction control result; the radio frequency power supply is controlled, the radio frequency power supply is controlled through the model predictive control algorithm model and the objective function constructed based on the Kalman filtering and the model predictive control algorithm, and the control efficiency of the radio frequency power supply is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radio frequency power supply control, in particular to a radio frequency power supply control method and device, electronic equipment and storage medium. BACKGROUND

[0002] Radio frequency power supply is widely used in semiconductor manufacturing, plasma processing and medical equipment fields, and the core goal of its control algorithm is to ensure the stability of output power, voltage or current.

[0003] The traditional control method mainly relies on PID algorithm control to realize output tracking through error signal adjustment, but this method has significant limitations. First, it has poor adaptability to nonlinear loads, and when the load impedance fluctuates, it is easy to cause system oscillation or steady-state error, affecting control accuracy. Second, the response speed is limited and cannot effectively respond to rapid load change scenarios such as mutations in the plasma ignition process, resulting in control lag. In addition, PID algorithm control cannot coordinate multi-variable coupling problems, such as the mutual influence between power, voltage and phase, often causing control imbalance. Finally, this method is sensitive to high-frequency noise, and the differential element may amplify the interference signal, further reducing system reliability. Although existing improved technologies such as adaptive proportional-integral-differential or fuzzy control can partially alleviate the above problems, they still have deficiencies in meeting fast response, strong robustness and multi-variable collaborative optimization, and cannot efficiently handle control requirements under complex working conditions.

[0004] Therefore, in order to solve the technical problems of poor adaptability, limited response speed, insufficient processing capacity for multi-variable coupling and sensitivity to high-frequency noise of the existing radio frequency power supply control method using PID algorithm, a radio frequency power supply control method, device, electronic equipment and storage medium are urgently needed. SUMMARY

[0005] The purpose of the present application is to provide a radio frequency power supply control method, device, electronic equipment and storage medium, which controls the radio frequency power supply by combining actual state data and a preset target function based on a model predictive control algorithm model constructed based on Kalman filtering and model predictive control algorithm, solves the problems of poor adaptability, limited response speed, insufficient processing capacity for multi-variable coupling and sensitivity to high-frequency noise of the existing radio frequency power supply control method using PID algorithm, makes the entire radio frequency power supply in a stable working state, ensures the radio frequency power supply in a good working state, and improves the control efficiency of the radio frequency power supply.

[0006] In a first aspect, the present application provides a radio frequency power supply control method for controlling a radio frequency power supply, comprising the steps of: obtaining the running parameters of the radio frequency power supply, and obtaining the historical running parameters and corresponding historical state data of the power supply database; construct a model predictive control algorithm model according to the historical operation parameters and corresponding historical state data through Kalman filtering and a model predictive control algorithm; input the operation parameters into the model predictive control algorithm model, and calculate to obtain predicted state data of the radio frequency power supply; based on a preset target function, combine the predicted state data and actual state data of the radio frequency power supply, and control the radio frequency power supply.

[0007] The radio frequency power supply control method provided in the application can control the radio frequency power supply, and through the model predictive control algorithm model constructed based on Kalman filtering and a model predictive control algorithm, the radio frequency power supply is controlled in combination with actual state data and a preset target function, thereby solving the problems of poor adaptability, limited response speed, insufficient processing capacity for multivariable coupling, and sensitivity to high-frequency noise of the existing radio frequency power supply control method using a PID algorithm, so that the entire radio frequency power supply is in a stable working state, the radio frequency power supply is ensured to be in a good working state, and the control efficiency of the radio frequency power supply is improved.

[0008] Optionally, the operation parameters include output voltage, output current, matching network capacitance value, PWM duty ratio, and capacitance adjustment amount of the radio frequency power supply; the historical operation parameters are historical data of the operation parameters; the historical state data are historical data of the state data; and the state data are output power and output voltage.

[0009] Optionally, the model predictive control algorithm model is constructed according to the historical operation parameters and corresponding historical state data through Kalman filtering and a model predictive control algorithm, including: construct a preliminary model predictive control algorithm model corresponding to the historical operation parameters and corresponding historical state data through Kalman filtering and a model predictive control algorithm; train the preliminary model predictive control algorithm model according to data for training in the historical operation parameters and corresponding historical state data, to obtain a trained preliminary model predictive control algorithm model; verify the trained preliminary model predictive control algorithm model based on data not used for training in the historical operation parameters and corresponding historical state data, to obtain the model predictive control algorithm model.

[0010] The radio frequency power supply control method provided in the application can control the radio frequency power supply, process noise and uncertainty through Kalman filtering, process multivariable prediction through a model predictive control algorithm, and construct and optimize a control model by using historical data, thereby improving the robustness and accuracy of radio frequency power supply control.

[0011] Optionally, the preliminary model predictive control algorithm model is trained according to the data for training in the historical operation parameters and the corresponding historical state data, to obtain a trained preliminary model predictive control algorithm model, including: The data for training in the historical operation parameters are input into the preliminary model predictive control algorithm model, to obtain corresponding output data; A training error is determined according to the data for training in the historical operation parameters, the corresponding state data and the corresponding output data; Based on the training error, parameters of the preliminary model predictive control algorithm model are adjusted to obtain optimal parameters, and the preliminary model predictive control algorithm model is optimized by using the optimal parameters, to obtain the trained preliminary model predictive control algorithm model.

[0012] Optionally, before the data for training in the historical operation parameters are input into the preliminary model predictive control algorithm model to obtain corresponding output data, the method further includes: Parameters of the preliminary model predictive control algorithm model are initialized.

[0013] Optionally, the predicted state data is output power and output voltage of a next time period.

[0014] Optionally, the radio frequency power source is controlled based on a preset target function, in combination with the predicted state data and actual state data of the radio frequency power source, including: Actual state data of the radio frequency power source are obtained; The predicted state data and the actual state data are input into the preset target function, to calculate an actual control amount of the radio frequency power source; The radio frequency power source is controlled based on the actual control amount.

[0015] The radio frequency power source control method provided by the application can control the radio frequency power source, the actual control amount calculated by the preset target function is used to control the radio frequency power source, so that the radio frequency power source is in a stable working state, and the control efficiency of the radio frequency power source is improved.

[0016] In a second aspect, the application provides a radio frequency power source control device for controlling a radio frequency power source, including: An acquisition module is configured to acquire operation parameters of the radio frequency power source, and acquire historical operation parameters and corresponding historical state data of a power source database; A construction module is configured to construct a model predictive control algorithm model according to the historical operation parameters and the corresponding historical state data by Kalman filtering and model predictive control algorithm; a calculation module, configured to input the operation parameter into the model predictive control algorithm model to calculate predicted state data of the radio frequency power supply; a control module, configured to control the radio frequency power supply based on a preset target function and in combination with the predicted state data and actual state data of the radio frequency power supply.

[0017] The radio frequency power supply control device controls the radio frequency power supply based on the model predictive control algorithm model constructed based on the Kalman filtering and the model predictive control algorithm, in combination with the actual state data and the preset target function, solves the problems of poor adaptability, limited response speed, insufficient processing capacity for multivariable coupling and sensitivity to high-frequency noise of the existing radio frequency power supply control method using the PID algorithm, makes the entire radio frequency power supply in a stable working state, ensures the radio frequency power supply in a good working state, and improves the control efficiency of the radio frequency power supply.

[0018] In a third aspect, the present application provides an electronic device including a processor and a memory, the memory storing a computer program executable by the processor, and the processor executes the computer program to run the steps of the radio frequency power supply control method as described above.

[0019] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and the computer program is executed by a processor to run the steps of the radio frequency power supply control method as described above. Advantages

[0020] The radio frequency power supply control method, device, electronic device and storage medium provided by the present application control the radio frequency power supply based on the model predictive control algorithm model constructed based on the Kalman filtering and the model predictive control algorithm, in combination with the actual state data and the preset target function, solve the problems of poor adaptability, limited response speed, insufficient processing capacity for multivariable coupling and sensitivity to high-frequency noise of the existing radio frequency power supply control method using the PID algorithm, make the entire radio frequency power supply in a stable working state, ensure the radio frequency power supply in a good working state, and improve the control efficiency of the radio frequency power supply. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The flowchart of the radio frequency power supply control method provided by the embodiment of the present application.

[0022] Figure 2 The structural schematic diagram of the radio frequency power supply control device provided by the embodiment of the present application.

[0023] Figure 3 The structural schematic diagram of the electronic device provided by the embodiment of the present application.

[0024] Label description: 1, acquisition module; 2, construction module; 3, calculation module; 4, control module; 301, processor; 302, memory; 303, communication bus. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0026] It should be noted that: similar labels and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0027] Please refer to Figure 1 , Figure 1 is a radio frequency power supply control method in some embodiments of the present application, used for controlling the radio frequency power supply, comprising: Step S101, obtaining the running parameters of the radio frequency power supply, and the historical running parameters and corresponding historical state data of the power supply database; Step S102, constructing a model predictive control algorithm model according to the historical running parameters and corresponding historical state data by Kalman filtering and model predictive control algorithm; Step S103, inputting the running parameters into the model predictive control algorithm model to calculate the predicted state data of the radio frequency power supply; Step S104, controlling the radio frequency power supply based on the preset target function, combining the predicted state data and the actual state data of the radio frequency power supply.

[0028] The radio frequency power supply control method controls the radio frequency power supply by combining the actual state data and the preset target function through the model predictive control algorithm model constructed based on the Kalman filtering and the model predictive control algorithm, solves the problems of poor adaptability, limited response speed, insufficient processing capacity for multivariable coupling and sensitivity to high-frequency noise of the existing radio frequency power supply control method when using the PID algorithm, makes the entire radio frequency power supply in a stable working state, ensures that the radio frequency power supply is in a good working state, and improves the control efficiency of the radio frequency power supply.

[0029] Specifically, in step S101, the operating parameters of the radio frequency power supply are obtained, including the output voltage, output current, matching network capacitance value, PWM duty ratio and capacitance adjustment amount of the radio frequency power supply.

[0030] The historical operating parameters and corresponding historical state data of the power supply database are obtained, wherein the historical operating parameters and corresponding historical state data include the historical records of the output voltage, output current, matching network capacitance value, PWM duty ratio and capacitance adjustment amount of the radio frequency power supply in the database (power supply database) at a certain time, and the historical state data corresponding to the historical records of these data, the corresponding historical state data includes the output power and output voltage at the next time (or next time period) after a certain time. Extract the historical operating parameters and corresponding historical state data of the same radio frequency power supply at each operating time in the database as the training set and the verification set, which can be used to train the model.

[0031] Specifically, in step S102, a model predictive control algorithm model is constructed according to the historical operating parameters and corresponding historical state data through Kalman filtering and model predictive control algorithm, including: A preliminary model predictive control algorithm model corresponding to the historical operating parameters and corresponding historical state data is constructed through Kalman filtering and model predictive control algorithm; The preliminary model predictive control algorithm model is trained according to the data for training in the historical operating parameters and corresponding historical state data, to obtain the trained preliminary model predictive control algorithm model; The trained preliminary model predictive control algorithm model is verified based on the data not used for training in the historical operating parameters and corresponding historical state data, to obtain the model predictive control algorithm model.

[0032] In step S102, a preliminary model predictive control algorithm model is constructed according to the historical operating parameters and corresponding historical state data by using Kalman filtering and model predictive control algorithm.

[0033] The specific calculation process of the model predictive control algorithm model is as follows: For the radio frequency power supply, the expression of the continuous-time state space model is specifically: ; ; wherein, x(k) is the state variable (i.e. output voltage, output current and matching network capacitance value) at the current time k, u(k) is the control input (i.e. PWM duty ratio and capacitance adjustment amount) at the current time k, x(k+1|k) is the predicted state variable at the time k+1, x(k) is the state data (i.e. output power and output voltage) at the time k, k is the time k, generally refers to the current time, k is a positive integer, A is the state transition matrix, B is the control input matrix, C is the observation matrix; w(k) is the process noise; v(k) is the measurement noise.

[0034] In the finite time domain from the current time k to the future, the state of the radio frequency power supply at the future time is predicted by using the continuous-time state space model, which can be expressed in the following form: x(k+i|k) = Ax(k+i−1|k) + Bu(k+i−1|k) ; y(k+i|k) = Cx(k+i|k) ; wherein, x(k+i|k) is the state variable at the time k+i predicted based on the data at the time k, x(k+i−1|k) is the state variable at the time k+i−1 predicted based on the data at the time k, i is the increment time length, i is a positive integer, i≤N p , N p is the prediction time domain, which describes the range of future time for evaluating the state of the radio frequency power supply at the future time in the optimization process, u(k+i−1|k) is the control input at the time k+i−1 predicted based on the data at the time k, y(k+i|k) is the state data at the time k+i predicted based on the data at the time k.

[0035] The Kalman filter is introduced to correct the state of the radio frequency power supply, and the corrected predicted state data is output: ; wherein, x(k|k) is the corrected state variable at the time k predicted based on the data at the time k; x(k−1|k−1) is the corrected state variable at the time k predicted based on the data at the time k−1; K is the Kalman gain, which can be calculated by the noise covariance and the existing Kalman gain calculation formula.

[0036] Thus, we get: ; wherein, is the state variable at the k+i moment based on the data prediction at the k moment.

[0037] The corrected state data at the future moment is calculated as the predicted state data output by the model predictive control algorithm, i.e. the predicted output power and output voltage of the next time period.

[0038] In the control problem of the radio frequency power supply, an optimization objective function (i.e. a preset objective function) is defined, which comprehensively considers the output error and the state constraint condition: ; u min ≤u(k+i|k)≤u max ; y min ≤y(k +i|k)≤y max ; wherein J is the objective function value, minJ represents the minimized objective function value; r(k+i) is the reference output (actual state data, i.e. actual output power and output voltage) at the k+i moment; is the control time domain, which describes the time length of the radio frequency power supply data at the future moment within the optimization interval, in order to ensure that the predicted state data can fully reflect the influence of the radio frequency power supply data at the future moment, it is necessary to satisfy ≤ , i≤ ; Q is a first weight matrix, and R is a second weight matrix; is the control amount change rate at the k+i moment based on the data prediction at the k moment; is the control input at the k+i moment based on the data prediction at the k moment; u min is the minimum value of the control input, and u max is the maximum value of the control input; y min is the minimum value of the state data, and y max is the maximum value of the state data; is the two-norm, and the minimized objective function value minJ corresponds to the numerical value of the control amount change rate as the actual control amount. Since the state constraint condition sets the upper and lower boundaries of the radio frequency power supply, i.e. sets the operating ranges of the PWM duty ratio, the capacitance adjustment amount, the output power and the output voltage, it can prevent the radio frequency power supply from being over-adjusted to cause hardware damage.

[0039] The quadratic programming (QP) algorithm is used to obtain the optimal control sequence u(k), u(k+1), …, u(k+Nc−1) as the actual control amount for the subsequent control steps.

[0040] Specifically, in step S102, a preliminary model predictive control algorithm model is trained according to historical running parameters and corresponding historical state data to obtain a trained preliminary model predictive control algorithm model, including: The data for training in the historical running parameters is input into the preliminary model predictive control algorithm model to obtain corresponding output data; The training error is determined according to the data for training in the historical running parameters, the corresponding historical state data, and the corresponding output data; Based on the training error, the parameters of the preliminary model predictive control algorithm model are adjusted to obtain optimal parameters, and the preliminary model predictive control algorithm model is optimized using the optimal parameters to obtain a trained preliminary model predictive control algorithm model.

[0041] In step S102, the training error is obtained by comparing the historical state data corresponding to the data for training in the historical running parameters in the database with the results obtained by inputting the data for training into the preliminary model predictive control algorithm model (the corresponding output data obtained by inputting the data for training in the historical running parameters into the preliminary model predictive control algorithm model). For example, the training error is the difference between part of the output data and the corresponding data in the historical state data, or the output data has one more relevant variable than the historical state data. The training error is used to adjust the parameters of the preliminary model predictive control algorithm model. For example, if the training error is the difference between part of the output data and the corresponding data in the historical state data, the corresponding model parameters are modified to make the output data more accurate, and optimal parameters are obtained. The preliminary model predictive control algorithm model is optimized using the optimal parameters to obtain a trained preliminary model predictive control algorithm model.

[0042] Although the preliminary model structure is determined when the preliminary model predictive control algorithm model is constructed, the values of the hyperparameters also have a significant impact on the performance of the model. Unlike the parameters of the model, the hyperparameters need to be manually set based on experience during the training process. Hyperparameters are usually closely related to the characteristics of the data, and appropriate hyperparameters can improve the accuracy and generalization ability of the model, while inappropriate selection may lead to overfitting or underfitting. However, manually adjusting hyperparameters requires a lot of manpower and time, and it is difficult to guarantee the best results. Therefore, the whale optimization algorithm (WOA) is used to automatically adjust the hyperparameters of the preliminary model predictive control algorithm model, so that the model better matches the characteristics of the running parameters, thereby improving the accuracy.

[0043] The principle of the whale optimization algorithm is as follows: WOA simulates the behavior of natural humpback whales searching for the optimal solution. The whole process mainly includes three stages: encircling prey, bubble-net attacking and searching for prey. Before using WOA to optimize the model parameters, the above three types of predatory behavior need to be mathematically modeled.

[0044] Encircling prey: Humpback whales can identify the location of prey and encircle it. In the process of encircling, humpback whales will select the search agent position closest to the prey as the optimal position, and other humpback whales will gradually approach the optimal position, and finally form an encirclement. This approach behavior is represented by the following mathematical formula: ; (1) ; (2) In the formula, N is the distance between the current whale individual and the optimal individual position; M * represents the position of the optimal whale individual in the current population; M * (t) is the position vector of the current whale individual; M (t) is the current optimal solution position vector; t is the current iteration number; E is the first coefficient vector; F is the second coefficient vector. The calculation formulas of the first coefficient vector E and the second coefficient vector F are as follows: ; (3) ; (4) In the formula, , t max is the maximum iteration number, a decreases linearly from 2 to 0 with the increase of iteration number; r is a random vector in [0, 1].

[0045] Bubble-net attacking method: Humpback whales use spiral motion to advance when hunting prey, and gradually reduce the encirclement. Among them, humpback whales choose to shrink the encirclement with a probability of P i , and choose to spiral forward with a probability of 1-P i . The mathematical model of this synchronous selection can be expressed as follows: ; (5) In the formula, P is a random number in [0, 1]; l is a random number in [-1, 1]; N' = |M*- M(t)| is the distance vector between the current whale individual and the optimal individual; b is a constant to describe the spiral shape; P iGenerally, 0.5 is taken. As the number of iterations increases, a will decrease linearly, while E fluctuates between [-a, a]; when E belongs to [-1, 1], the whale launches an attack on the prey, and the next position of the whale individual is an arbitrary position between the current position and the prey.

[0046] Search for prey: In the search for prey, the search mechanism of the whale individual depends on the size of |E|. When |E| < 1, the whale individual will perform a local search according to the position update of the optimal position. When |E| > 1, the whale individual will move away from the current optimal position and perform a global search according to the position update of the randomly selected whale position. When |E| > 1, the mathematical model of global search is as follows: ; (6) ; (7) In the formula, M rand (t) is the position vector of the randomly selected whale individual.

[0047] For the preliminary model predictive control algorithm model, adjust the control horizon , the first weight matrix Q and the second weight matrix R, etc. Three key hyperparameters can have a significant impact on model performance. When training a deep learning model, the control horizon controls the step size of each parameter update. The first weight matrix Q is the output tracking weight, which is used to balance the system tracking performance to ensure the stability of the plasma process power. The second weight matrix R is the control amount change rate weight, which is used to control the action cost to extend the matching network capacitor life and reduce the hardware failure rate. In addition to the above three key hyperparameters, according to experience, the batch size during training is set to 64, and since the preliminary model predictive control algorithm model has a small capacity, a total of 50 rounds of iterations are performed during training.

[0048] The control horizon , the first weight matrix Q and the second weight matrix R are used as the hyperparameters to be optimized, and the upper and lower bounds of the search range are set to the hyperparameters to be optimized. The hyperparameters to be optimized are initialized to determine the initial whale population position M(Q, R, ). Calculate the error rate after each iteration while iterating the hyperparameters to be optimized. When the number of iterations reaches the maximum number of iterations T max , stop the iteration, extract the minimum value from the error rate after each iteration, and determine the hyperparameters after the iteration corresponding to the minimum value as the optimal hyperparameters.

[0049] In the WOA optimization process, the prey position is assumed to be the optimal solution, and the position of each whale is regarded as a potential solution. In each iteration, the position update strategy of each whale is determined according to the value of the random number P and the modulus of the coefficient vector A. As the iteration proceeds, the whale population gradually approaches the optimal solution. The specific steps of the model training and optimization process of the preliminary model predictive control algorithm are as follows: Step 1: Data set preparation and division. The data used for training in the historical running parameters are labeled according to different wear stages, and divided into training data (used to optimize model parameters) and test data (used to optimize model hyperparameters) according to a certain proportion. Finally, the data samples are normalized and input into the preliminary model predictive control algorithm model.

[0050] Step 2: WOA parameter initialization. The control time domain , the first weight matrix Q and the second weight matrix R are set as the optimization parameters (to be optimized hyperparameters), and the upper and lower bounds of the search range are set. The number of whale populations Z, the maximum number of iterations Tmax and the spatial dimension of the whale individual position are set, and the whale population position M(Q, R, ) is initialized.

[0051] Step 3: Fitness evaluation. The network model is trained and tested, and then the fitness (objective function value) of each whale individual in the population is calculated. The position of the whale individual with the smallest fitness value is taken as the current optimal position and is recorded, i.e. the global optimal solution is determined. The recognition error rate of the network model during testing is taken as the objective function of WOA, as shown in the following formula: ; (8) In the formula, min Error is the minimization objective function, Ture Num is the number of correctly classified samples in the test data; Total Num is the total number of samples in the test data; In the iteration process, WOA obtains the optimal set of hyperparameters by minimizing the objective function.

[0052] Step 4: Update individual position. A random number P is generated. When P≥0.5, the whale individual position is updated in a spiral advancing manner according to formula (5); when P<0.5 and |E|≥1, the whale individual position is updated in a shrinking surrounding manner according to formula (2); when P<0.5 and |E|<1, the whale individual position is updated in a global search manner according to formula (7).

[0053] Step 5: Iteration process judgment. Determine whether the termination condition (whether the number of iterations reaches the maximum number of iterations Tmax max), if the termination iteration condition is met, the optimal hyperparameters are output, otherwise step 3 is returned to continue execution until the termination condition is met.

[0054] Step 6: Save the model. Save the optimal hyperparameters, and use the optimal hyperparameters to optimize the trained preliminary model predictive control algorithm model.

[0055] In step S102, the historical operation parameters not used for training in the database (data not used for training in the historical operation parameters) are input into the trained preliminary model predictive control algorithm model to obtain the output data output by the preliminary model predictive control algorithm model as state data verification data. The state data verification data is compared with the historical state data corresponding to the historical operation parameters not used for training in the database to determine whether the error is within an acceptable range (the error acceptable range is generally 0 to 3%, which can be modified according to actual needs). The accuracy of the trained preliminary model predictive control algorithm model is verified, and the model predictive control algorithm model is obtained.

[0056] Specifically, before the historical operation parameters are input into the preliminary model predictive control algorithm model to obtain the corresponding output data in step S102, the following steps are further included: Initialize the parameters of the preliminary model predictive control algorithm model.

[0057] In step S102, the parameters of the preliminary model predictive control algorithm model are initialized before the input data is used to train the model, so as to ensure that the model is in a normal use state.

[0058] Specifically, in step S103, the operation parameters of the radio frequency power supply are input into the model predictive control algorithm model, and the predicted state data of the radio frequency power supply is obtained through the calculation of the model predictive control algorithm model.

[0059] Specifically, in step S104, based on the preset target function, the radio frequency power supply is controlled in combination with the predicted state data and the actual state data of the radio frequency power supply, including: Obtain the actual state data of the radio frequency power supply; Input the predicted state data and the actual state data into the preset target function to calculate the actual control amount of the radio frequency power supply; Control the radio frequency power supply based on the actual control amount.

[0060] In step S104, the actual state data of the radio frequency power supply is acquired, the predicted state data and the actual state data at the corresponding moment are input into a preset target function (i.e. an optimization target function constructed when the model predictive control algorithm model is constructed), and the actual control amount of the radio frequency power supply, i.e. the control amount change rate corresponding to the PWM duty ratio and the capacitance adjustment amount, is calculated under the state constraint condition, i.e. within the operating range of the PWM duty ratio, the capacitance adjustment amount, the output power and the output voltage. Based on the actual control amount, the radio frequency power supply is controlled, the actual control amount is taken as the adjustment amount, the adjusted PWM duty ratio and capacitance adjustment amount tend to the control input corresponding to the actual control amount, and the PWM duty ratio, the capacitance adjustment amount, the output power and the output voltage are always within the operating range for the radio frequency power supply to work normally due to the limitation of the state constraint condition, so that the target function is at a minimum value and the radio frequency power supply is in a stable working state, thereby ensuring that the radio frequency power supply is in a good working state.

[0061] As can be seen from the above, the radio frequency power supply control method acquires the operating parameters of the radio frequency power supply and the historical operating parameters and corresponding historical state data of the power supply database, constructs a model predictive control algorithm model according to the historical operating parameters and corresponding historical state data through Kalman filtering and model predictive control algorithm, inputs the operating parameters into the model predictive control algorithm model, calculates the predicted state data of the radio frequency power supply, controls the radio frequency power supply based on the preset target function in combination with the predicted state data and the actual state data of the radio frequency power supply; thereby, the model predictive control algorithm model constructed based on Kalman filtering and model predictive control algorithm is used to control the radio frequency power supply in combination with the actual state data and the preset target function, the problems of poor adaptability, limited response speed, insufficient processing capacity for multivariable coupling and sensitivity to high-frequency noise of the existing radio frequency power supply control method using PID algorithm are solved, the entire radio frequency power supply is in a stable working state, the radio frequency power supply is ensured to be in a good working state, and the control efficiency of the radio frequency power supply is improved.

[0062] Reference Figure 2 The present application provides a radio frequency power supply control device for controlling a radio frequency power supply, comprising: The acquisition module 1 is used to acquire the operating parameters of the radio frequency power supply and the historical operating parameters and corresponding historical state data of the power supply database; The construction module 2 is used to construct a model predictive control algorithm model according to the historical operating parameters and corresponding historical state data through Kalman filtering and model predictive control algorithm; The calculation module 3 is used to input the operating parameters into the model predictive control algorithm model to calculate the predicted state data of the radio frequency power supply; The control module 4 is configured to control the radio frequency power supply based on a preset target function, in combination with the predicted state data and the actual state data of the radio frequency power supply.

[0063] The radio frequency power supply control device controls the radio frequency power supply based on the model predictive control algorithm model constructed based on the Kalman filtering and the model predictive control algorithm, in combination with the actual state data and the preset target function, solves the problems of poor adaptability, limited response speed, insufficient processing capacity for multivariable coupling, and sensitivity to high-frequency noise when the existing radio frequency power supply control method uses the PID algorithm, enables the entire radio frequency power supply to be in a stable working state, ensures that the radio frequency power supply is in a good working state, and improves the control efficiency of the radio frequency power supply.

[0064] Specifically, when the acquisition module 1 is executed, the running parameters of the radio frequency power supply are acquired, and the running parameters include the output voltage, the output current, the matching network capacitance value, the PWM duty ratio, and the capacitance adjustment amount of the radio frequency power supply.

[0065] The historical running parameters and corresponding historical state data of the power supply database are acquired, wherein the historical running parameters and corresponding historical state data include the historical records of the output voltage, the output current, the matching network capacitance value, the PWM duty ratio, and the capacitance adjustment amount of the radio frequency power supply in the database (power supply database) at a certain time, and the historical state data corresponding to the historical records of these data, and the corresponding historical state data includes the output power and the output voltage at the next time (or the next time period) after a certain time. The historical running parameters and corresponding historical state data of each running time of the same radio frequency power supply in the database are extracted as a training set and a verification set, which can be used to train the model.

[0066] Specifically, when the construction module 2 constructs the model predictive control algorithm model according to the historical running parameters and corresponding historical state data by using the Kalman filtering and the model predictive control algorithm, the following is performed: A preliminary model predictive control algorithm model corresponding to the historical running parameters and corresponding historical state data is constructed by using the Kalman filtering and the model predictive control algorithm; The preliminary model predictive control algorithm model is trained according to the data for training in the historical running parameters and corresponding historical state data, to obtain a trained preliminary model predictive control algorithm model; The trained preliminary model predictive control algorithm model is verified based on the data not used for training in the historical running parameters and corresponding historical state data, to obtain the model predictive control algorithm model.

[0067] When the construction module 2 is executed, a preliminary model predictive control algorithm model is constructed according to the historical running parameters and corresponding historical state data by using the Kalman filtering and the model predictive control algorithm.

[0068] The specific calculation process of the model predictive control algorithm model is as follows: For the radio frequency power supply, the expression of the continuous-time state space model is as follows: ; ; wherein, x(k) is the state variable (i.e., the output voltage, the output current, and the matching network capacitance value) at the current time k, u(k) is the control input (i.e., the PWM duty ratio and the capacitance adjustment amount) at the current time k, x(k+1|k) is the predicted state variable at the (k+1)th time, y(k) is the state data (i.e., the output power and the output voltage) at the kth time, k is the kth time, generally referring to the current time, k is a positive integer, A is the state transition matrix, B is the control input matrix, and C is the observation matrix; w(k) is the process noise; v(k) is the measurement noise.

[0069] Within a limited time domain from the current time k to the future, the radio frequency power supply state at the future time is predicted using the continuous-time state space model, which can be expressed in the following form: x(k+i|k) = Ax(k+i−1|k) + Bu(k+i−1|k) ; y(k+i|k) = Cx(k+i|k) ; wherein, x(k+i|k) is the state variable at the (k+i)th time predicted based on the data at the kth time, x(k+i−1|k) is the state variable at the (k+i−1)th time predicted based on the data at the kth time, i is the incremental time length, i is a positive integer, i≤N p , N p is the prediction time domain, describing the range of future times for evaluating the radio frequency power supply state at the future times during the optimization process, u(k+i−1|k) is the control input at the (k+i−1)th time predicted based on the data at the kth time, and y(k+i|k) is the state data at the (k+i)th time predicted based on the data at the kth time.

[0070] The Kalman filter is introduced to correct the radio frequency power supply state, and the corrected predicted state data is output: ; wherein, x(k|k) is the corrected state variable at the kth time predicted based on the data at the kth time; is the state variable at the kth moment predicted based on the data at the k-1th moment; K is the Kalman gain, which can be calculated by the noise covariance and the existing Kalman gain calculation formula.

[0071] Thus, the following is obtained: ; wherein, is the state variable at the k+i th moment predicted based on the data at the kth moment.

[0072] The corrected future moment state data is calculated as the predicted state data output by the model predictive control algorithm model, that is, the predicted output power and output voltage of the next time period.

[0073] In the control problem of the radio frequency power supply, an optimization objective function (i.e., a preset objective function) is defined, and the output error and the state constraint condition are comprehensively considered: ; u min ≤u(k+i|k)≤u max ; y min ≤y(k +i|k)≤y max ; wherein, J is the objective function value, minJ represents the minimized objective function value; r(k+i) is the reference output (actual state data, i.e., actual output power and output voltage) at the k+i th moment; is the control time domain, which describes the time length of the radio frequency power supply data at the future moment within the optimization interval, and in order to ensure that the predicted state data can fully reflect the influence of the radio frequency power supply data at the future moment, the following condition needs to be met: ≤ , i≤ ; Q is the first weight matrix, and R is the second weight matrix; is the control amount change rate at the k+i th moment predicted based on the data at the kth moment; is the control input at the k+i th moment predicted based on the data at the kth moment; u min is the minimum value of the control input, and u max is the maximum value of the control input; y min is the minimum value of the state data, and y max is the maximum value of the state data; is the two-norm, and the control amount change rate corresponding to the minimized objective function value minJ can be calculated by the two-norm. The numerical value of the state constraint is taken as the actual control amount. Since the state constraint sets the upper and lower boundaries of the radio frequency power supply, i.e., sets the operating range of the PWM duty ratio, the capacitance adjustment amount, the output power and the output voltage, it can prevent the radio frequency power supply from overshooting and causing hardware damage.

[0074] The optimal control sequence u(k), u(k+1), …, u(k+Nc-1) is obtained by using a quadratic programming (QP) algorithm and is taken as the actual control amount for subsequent control steps.

[0075] Specifically, when the training module 2 trains the preliminary model predictive control algorithm model according to the historical operating parameters and the corresponding historical state data to obtain the trained preliminary model predictive control algorithm model, the training module 2 performs the following operations: The data for training in the historical operating parameters is input into the preliminary model predictive control algorithm model to obtain corresponding output data; The training error is determined according to the data for training in the historical operating parameters, the corresponding historical state data and the corresponding output data; Based on the training error, the parameters of the preliminary model predictive control algorithm model are adjusted to obtain optimal parameters, and the preliminary model predictive control algorithm model is optimized by using the optimal parameters to obtain the trained preliminary model predictive control algorithm model.

[0076] When the training module 2 is executed, the historical state data corresponding to the data for training in the historical operating parameters in the database is compared with the result obtained by inputting the data for training into the preliminary model predictive control algorithm model (the corresponding output data obtained by inputting the data for training into the preliminary model predictive control algorithm model), and the training error is obtained, such as the difference between part of the output data and the corresponding data in the historical state data or the training error that the output data has one more relevant variable than the historical state data. The training error is used to adjust the parameters of the preliminary model predictive control algorithm model. For example, if the training error is the difference between part of the output data and the corresponding data in the historical state data, the corresponding model parameters are modified to make the output data more accurate, and the optimal parameters are obtained. The preliminary model predictive control algorithm model is optimized by using the optimal parameters to obtain the trained preliminary model predictive control algorithm model.

[0077] Although the preliminary model structure is determined when constructing the preliminary model predictive control algorithm model, the values of hyperparameters also have a significant impact on the performance of the model. Unlike the parameters of the model, hyperparameters need to be manually set according to experience during the training process. Hyperparameters are usually closely related to the characteristics of the data, and appropriate hyperparameters can improve the accuracy and generalization ability of the model, while improper selection may lead to overfitting or underfitting. However, manually adjusting hyperparameters requires a lot of manpower and time, and it is difficult to guarantee the best effect. Therefore, the whale optimization algorithm (WOA) is used to automatically adjust the hyperparameters of the preliminary model predictive control algorithm model, so that the model better matches the operating parameter characteristics, thereby improving the accuracy.

[0078] The principle of the whale optimization algorithm is as follows: WOA simulates the behavior of humpback whales searching for prey to obtain the optimal solution. The entire process of the algorithm mainly includes three stages: encircling prey, bubble-net attacking, and search for prey. Before using WOA to optimize the model parameters, the above three types of predatory behavior need to be mathematically modeled.

[0079] Encircling prey: Humpback whales can identify the location of prey and encircle it. During the encircling process, humpback whales will select the search agent position closest to the prey as the optimal position, and other humpback whales will gradually approach the optimal position, eventually forming an encirclement. This approximation behavior is represented by the following mathematical formula: ; (1) ; (2) In the formula, N is the distance between the current whale individual and the optimal individual position; M * represents the position of the optimal whale individual in the current population; M * (t) is the current optimal solution position vector; t is the current iteration number; E is the first coefficient vector; F is the second coefficient vector. The calculation formulas of the first coefficient vector E and the second coefficient vector F are as follows: ; (3) ; (4) In the formula, , t max is the maximum iteration number, a decreases linearly from 2 to 0 as the iteration number increases; r is a random vector in [0, 1].

[0080] Bubble-net attacking method: The humpback whale adopts a spiral motion to advance and gradually narrows the encirclement when hunting prey. The humpback whale chooses to shrink the encirclement with a probability of P i and chooses to advance in a spiral with a probability of 1-P i . The mathematical model of this synchronous selection can be expressed as follows: ; (5) where P is a random number in [0, 1]; l is a random number in [-1, 1]; N' = |M*- M(t)| is the distance vector between the current whale individual and the optimal individual; b is a constant used to describe the spiral shape; P i is generally taken as 0.5. As the number of iterations increases, a decreases linearly, and at the same time, E fluctuates between [-a, a]; when E belongs to [-1, 1], the whale launches an attack on the prey, and the next position of the whale individual is any position between the current position and the prey.

[0081] Search for prey: When searching for prey, the search mechanism of the whale individual depends on the size of |E|. When |E| < 1, the whale individual will update its position according to the optimal position, performing local search. When |E| > 1, the whale individual will move away from the current optimal position and update its position according to a randomly selected whale position, performing global search. When |E| > 1, the mathematical model of global search is as follows: ; (6) ; (7) where M rand (t) is the position vector of the randomly selected whale individual.

[0082] For the preliminary model predictive control algorithm model, adjusting three key hyperparameters such as control horizon , the first weight matrix Q and the second weight matrix R can have a significant impact on the model performance. When training a deep learning model, the control horizon controls the step size of each parameter update. The first weight matrix Q is the output tracking weight, which is used to balance the system tracking performance to ensure the stability of the plasma process power. The second weight matrix R is the control amount change rate weight, which is used to control the action cost to extend the matching network capacitor life and reduce the hardware failure rate. In addition to the above three key hyperparameters, according to experience, the batch size during training is set to 64, and since the preliminary model predictive control algorithm model has a small capacity, a total of 50 rounds of iterations are performed during the training process.

[0083] The control horizon , the first weight matrix Q and the second weight matrix R are taken as the to-be-optimized hyperparameters, and the upper and lower boundaries of the search range are set to set the to-be-optimized hyperparameters. The to-be-optimized hyperparameters are initialized to determine the initial whale population position M(Q, R, ). max The error rate after each iteration is calculated while iterating the to-be-optimized hyperparameters, and after the number of iterations reaches the maximum number of iterations T max , the iteration is stopped, the minimum value is extracted from the error rate after each iteration, and the hyperparameters after the iteration corresponding to the minimum value are determined as the optimal hyperparameters.

[0084] In the WOA optimization process, the prey position is assumed to be the optimal solution, and the position of each whale is regarded as a potential solution. In each iteration, the position update strategy of each whale is determined according to the value of the random number P and the modulus of the coefficient vector A. As the iteration proceeds, the whale population gradually approaches the optimal solution. The specific steps of the model training and optimization process of the preliminary model predictive control algorithm are as follows: Step 1: Data set preparation and division. The data used for training in the historical running parameters are labeled according to different wear stages, and divided into training data (used to optimize model parameters) and test data (used to optimize model hyperparameters) according to a certain proportion. Finally, the data samples are normalized and input into the preliminary model predictive control algorithm model.

[0085] Step 2: WOA parameter initialization. The control time domain , the first weight matrix Q and the second weight matrix R are taken as the to-be-optimized hyperparameters, and the upper and lower boundaries of the search range are set. The number of whale populations Z, the maximum number of iterations Tmax, and the spatial dimension of the whale individual position are set, and the whale population position M(Q, R, ) is initialized.

[0086] Step 3: Fitness evaluation. The network model is trained and tested, and then the fitness (objective function value) of each whale individual in the population is calculated. The position of the whale individual with the minimum fitness value is taken as the current optimal position and is recorded, i.e. the global optimal solution is determined. The recognition error rate of the network model during testing is taken as the objective function of WOA, as shown in the following formula: ; (8) In the formula, min Error is the minimization objective function, Ture Num is the number of correctly classified samples in the test data; Total Num is the total number of samples in the test data; In the iteration process, WOA obtains the optimal set of hyperparameters by minimizing the objective function.

[0087] Step 4: updating the individual position. A random number P is generated, when P≥0.5, the whale individual position is updated in a spiral advancing manner according to formula (5); when P<0.5 and |E|≥1, the whale individual position is updated in a shrinking enclosing manner according to formula (2); when P<0.5 and |E|<1, the whale individual position is updated in a global search manner according to formula (7).

[0088] Step 5: iteration process judgment. It is judged whether the termination condition (whether the maximum iteration number T max ) is reached, if the termination iteration condition is met, the optimal hyperparameter is output, otherwise step 3 is returned to continue execution until the termination condition is met.

[0089] Step 6: saving the model. The optimal hyperparameter is saved, and the optimal hyperparameter is used to optimize the preliminary model predictive control algorithm model after training.

[0090] When the construction module 2 is executed, the historical running parameters in the database that are not used for training (the data in the historical running parameters that are not used for training) are input into the preliminary model predictive control algorithm model after training, and the output data output by the preliminary model predictive control algorithm model is obtained as state data verification data. By comparing the state data verification data with the historical state data corresponding to the historical running parameters not used for training in the database, it is determined that the error is within an acceptable range (the error acceptable range is generally 0 to 3%, which can be modified according to actual needs), the accuracy of the preliminary model predictive control algorithm model after training is verified, and the model predictive control algorithm model is obtained.

[0091] Specifically, before the construction module 2 inputs the historical running parameters into the preliminary model predictive control algorithm model to obtain the corresponding output data, it also performs: initializing the parameters of the preliminary model predictive control algorithm model.

[0092] When the construction module 2 is executed, the parameters of the preliminary model predictive control algorithm model are initialized before the input data is trained on the model to ensure that the model is in a normal use state.

[0093] Specifically, when the calculation module 3 is executed, the running parameters of the radio frequency power supply are input into the model predictive control algorithm model, and the predicted state data of the radio frequency power supply is obtained through the calculation of the model predictive control algorithm model.

[0094] Specifically, when the control module 4 controls the radio frequency power supply based on the preset target function, combines the predicted state data and the actual state data of the radio frequency power supply, and performs: obtaining the actual state data of the radio frequency power supply; The predicted state data and the actual state data are input into a preset target function, and an actual control amount of the radio frequency power supply is calculated; The radio frequency power supply is controlled based on the actual control amount.

[0095] When the control module 4 is executed, the actual state data of the radio frequency power supply is acquired, the predicted state data and the actual state data at the corresponding time are input into a preset target function (i.e. the optimization target function constructed when the model predictive control algorithm model is constructed), and the actual control amount of the radio frequency power supply, i.e. the control amount change rate corresponding to the PWM duty ratio and the capacitance adjustment amount, is calculated under the state constraint condition, i.e. within the operating range of the PWM duty ratio, the capacitance adjustment amount, the output power and the output voltage. The radio frequency power supply is controlled based on the actual control amount, the actual control amount is used as the adjustment amount, the adjusted PWM duty ratio and capacitance adjustment amount tend to the control input corresponding to the actual control amount, and the PWM duty ratio, the capacitance adjustment amount, the output power and the output voltage are always within the operating range that enables the radio frequency power supply to work normally due to the limitation of the state constraint condition, so as to minimize the target function and keep the radio frequency power supply in a stable working state, thereby ensuring that the radio frequency power supply is in a good working state.

[0096] As can be seen from the above, the radio frequency power supply control device acquires the operating parameters of the radio frequency power supply and the historical operating parameters and corresponding historical state data of the power supply database, constructs a model predictive control algorithm model according to the historical operating parameters and corresponding historical state data through Kalman filtering and model predictive control algorithm, inputs the operating parameters into the model predictive control algorithm model, calculates the predicted state data of the radio frequency power supply, controls the radio frequency power supply based on the preset target function and the predicted state data and the actual state data of the radio frequency power supply. Thus, the model predictive control algorithm model constructed based on Kalman filtering and model predictive control algorithm is used to control the radio frequency power supply in combination with the actual state data and the preset target function, which solves the problems of poor adaptability, limited response speed, insufficient processing capacity for multivariable coupling and sensitivity to high-frequency noise of the existing radio frequency power supply control method using PID algorithm, keeps the entire radio frequency power supply in a stable working state, ensures that the radio frequency power supply is in a good working state, and improves the control efficiency of the radio frequency power supply.

[0097] Please refer to Figure 3 , Figure 3A structural schematic diagram of an electronic device provided by an embodiment of the present application is provided, and the present application provides an electronic device, comprising: a processor 301 and a memory 302, the processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanism (not marked), the memory 302 stores a computer program executable by the processor 301, when the electronic device is running, the processor 301 executes the computer program to execute the radio frequency power supply control method in any optional implementation manner of the above-mentioned embodiments, to realize the following functions: obtaining the running parameter of the radio frequency power supply, and obtaining the historical running parameter and the corresponding historical state data of the power supply database, constructing the model predictive control algorithm model according to the historical running parameter and the corresponding historical state data through the Kalman filtering and the model predictive control algorithm, inputting the running parameter into the model predictive control algorithm model, calculating the predicted state data of the radio frequency power supply, and controlling the radio frequency power supply based on the preset target function and in combination with the predicted state data and the actual state data of the radio frequency power supply.

[0098] The present application provides a computer readable storage medium, which stores a computer program, when the computer program is executed by a processor, the radio frequency power supply control method in any optional implementation manner of the above-mentioned embodiments is executed, to realize the following functions: obtaining the running parameter of the radio frequency power supply, and obtaining the historical running parameter and the corresponding historical state data of the power supply database, constructing the model predictive control algorithm model according to the historical running parameter and the corresponding historical state data through the Kalman filtering and the model predictive control algorithm, inputting the running parameter into the model predictive control algorithm model, calculating the predicted state data of the radio frequency power supply, and controlling the radio frequency power supply based on the preset target function and in combination with the predicted state data and the actual state data of the radio frequency power supply. The storage medium can be realized by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0099] In the embodiments of the present application, it should be understood that the disclosed apparatus and method can be implemented in other manners. The embodiments described above are merely exemplary, for example, the division of the units is only a logical function division, and there can be another division manner in actual implementation; for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.

[0100] In addition, the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, and can be located in one position, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purposes of the embodiments of the present application.

[0101] In addition, the functional modules in the various embodiments of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0102] In this article, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0103] The above only describes the embodiments of the present application, and is not used to limit the protection scope of the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A radio frequency power supply control method for controlling a radio frequency power supply, characterized in that: Including steps: Obtaining the operating parameters of the RF power supply, as well as the historical operating parameters and corresponding historical status data from the power supply database; Constructing a model predictive control algorithm model based on the historical operating parameters and corresponding historical state data through Kalman filtering and model predictive control algorithm; Inputting the operating parameters into the model predictive control algorithm model to calculate predicted state data of the radio frequency power supply; Based on a preset objective function, the radio frequency power supply is controlled in combination with the predicted state data and the actual state data of the radio frequency power supply.

2. The radio frequency power supply control method according to claim 1, wherein: The operating parameters include the output voltage, output current, matching network capacitance value, PWM duty cycle and capacitance adjustment amount of the RF power supply; the historical operating parameters are historical data of the operating parameters; the historical status data are historical data of the status data; the status data are output power and output voltage.

3. The radio frequency power supply control method according to claim 2, wherein: By using Kalman filtering and a model predictive control algorithm, a model predictive control algorithm model is constructed based on the historical operating parameters and the corresponding historical state data, including: Constructing a preliminary model predictive control algorithm model corresponding to the historical operating parameters and the corresponding historical state data through Kalman filtering and model predictive control algorithm; Training the preliminary model predictive control algorithm model based on the historical operating parameters and the corresponding historical state data for training, to obtain a trained preliminary model predictive control algorithm model; Based on the historical operating parameters and the data not used for training in the corresponding historical state data, the trained preliminary model predictive control algorithm model is verified to obtain the model predictive control algorithm model.

4. The radio frequency power supply control method according to claim 3, wherein: Training the preliminary model predictive control algorithm model based on the historical operating parameters and the corresponding historical state data for training, to obtain a trained preliminary model predictive control algorithm model, including: Inputting the data used for training in the historical operating parameters into the preliminary model predictive control algorithm model to obtain corresponding output data; Determining a training error based on the data used for training in the historical operating parameters, the corresponding state data, and the corresponding output data; Based on the training error, the parameters of the preliminary model predictive control algorithm model are adjusted to obtain the optimal parameters, and the preliminary model predictive control algorithm model is optimized using the optimal parameters to obtain the trained preliminary model predictive control algorithm model.

5. The radio frequency power supply control method according to claim 4, characterized in that: Before inputting the training data of the historical operating parameters into the preliminary model predictive control algorithm model to obtain the corresponding output data, the method further includes: Initialize the parameters of the preliminary model predictive control algorithm model.

6. The radio frequency power supply control method according to claim 1, wherein: The predicted state data is the output power and output voltage predicted for the next time period.

7. The radio frequency power supply control method according to claim 1, wherein: Based on a preset objective function, the radio frequency power supply is controlled in combination with the predicted state data and the actual state data of the radio frequency power supply, including: Acquiring actual status data of the radio frequency power supply; Inputting the predicted state data and the actual state data into a preset objective function to calculate the actual control amount of the radio frequency power supply; The radio frequency power supply is controlled based on the actual control variable.

8. A radio frequency power supply control device for controlling a radio frequency power supply, characterized in that: include: An acquisition module is used to obtain the operating parameters of the RF power supply, as well as the historical operating parameters and corresponding historical status data of the power supply database; A construction module, configured to construct a model predictive control algorithm model based on the historical operating parameters and corresponding historical state data by using a Kalman filter and a model predictive control algorithm; a calculation module, configured to input the operating parameters into the model predictive control algorithm model and calculate predicted state data of the radio frequency power supply; A control module is used to control the radio frequency power supply based on a preset objective function in combination with the predicted state data and the actual state data of the radio frequency power supply.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, the method runs the steps of the radio frequency power supply control method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the radio frequency power supply control method according to any one of claims 1 to 7 are executed.

Images